Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Boris Kment
University of Michigan, Ann Arbor
This paper is, in part, a straightforward exercise in philosophical analysis: I
will try to define metaphysical necessity. But I will combine this aim with another:
I want to know which cognitive practices of ordinary life gave rise to the concept
of necessity, and what role the notion plays in these practices. Let me describe
this goal in more detail, before giving an overview of my strategy.
which ordinary-life practices give rise to modal notions, and what role modal
concepts play in them. It should thereby elucidate what the purpose of these
notions is, why creatures with our interests and concerns have developed them.
In this way, a comprehensive theory of modality should combine a theory of
necessity with a theory of the practice of modalizing. The point can be explained
by the metaphor of the difference between an internal and an external viewpoint.
Before we start to philosophize about necessity, we have an implicit theory
about it. The philosopher provides this pre-philosophical system of beliefs about
modality with a foundation, and refines, extends and corrects it from within. He
acts as a participant in our practice of modalizing; his standpoint is internal to
this practice. But the philosopher should also make the practice of modalizing
itself an object of study. He should, as it were, take a standpoint external to the
practice, in order to describe the practice, and explain what its function is, why it
exists.
It is a familiar fact that these two kinds of interest can pull in opposite
directions. If we concentrate exclusively on the task of giving a metaphysical
account of what necessity is, we might end up with a theory that makes it
hard to explain why we are interested in modality. It is a common charge
whether it is justified or not I shall not endeavor to decideagainst Lewiss
account of necessity that it fails in just this way.1 Philosophers have objected
that, even if there were other worlds in Lewiss sense, we would have no apparent
reason to be interested in what goes on in them. Hence, Lewiss account, so the
objection continues, makes it a mystery why we should bother to think about
modal facts. At the other end of the spectrum there are theories that do well
at explaining the purpose of our modal notions, but do so at the expense of
giving implausible accounts of what necessity is. Consider the conventionalist
theory that for a proposition to be necessary is for it to owe its truth to a
convention, perhaps the convention that we ought to regard the proposition as
true come what may. If some propositions are indeed conventionally exempted
from empirical testing, then it benefits our epistemic practices if we possess
a concept that singles them out. The conventionalist therefore has no great
difficulty with explaining the point of our modal notions. But this advantage
is purchased at the price of several well-known drawbacks in her theory of
necessity.
The task is thus to develop an approach to modality that permits us to
achieve both of our goals, a credible metaphysical theory of necessity and a
plausible account of the practice of modalizing. I think that the best way of doing
this is not to neglect either objective while pursuing the other, but to integrate
the two goals in a single enterprise: an account of the nature of metaphysical
necessity can be guided by a hypothesis about the ordinary-life practice in which
the concept of necessity originated, while assumptions about what necessity is can
in turn suggest ways of developing ones ideas about the practice of modalizing.
This is the methodology I will employ.
natural to regard the necessity operator as expressing the most basic concept of
modal discourse, and to define other modal notions (like that of a possible world)
in terms of modal operators.3,4
I think that both intuitions, the otherworldliness intuition and the modalist
intuition, are very forceful, and I believe that a plausible metaphysical account of
modality needs to be capable of accommodating both intuitions. It is important
to be clear about what is required in order to accommodate them. I take both
the otherworldliness intuition and the modalist intuition to be intuitions about
what it is for a proposition to be necessary. According to the first intuition, to
be necessary just is to be true in all scenarios that are alternatives to the way
things actually are; according to the modalist intuition, it is to be true in some
particularly secure and inexorable way. Now suppose that we propose, as our
account of necessity, that to be necessary is to be F (for some predicate F ). In
order for this view to accommodate the two intuitions it is not sufficient that it
can be shown that the propositions that have the property of F-ness are just those
that are true in all alternatives to the way things are, or that the propositions that
are F are just those whose truth is particularly secure in the sense underlying the
modalist intuition. Rather, the property of F-ness must be (identical with) the
property of being true in all alternatives to the way things are, and it must be
(identical with) the property of being true in a particularly secure way. In order
to accommodate both intuitions, we need to find some way of interpreting the
phrases true in all alternatives to the way things are and true in a particularly
inexorable way on which the two phrases single out the same property, and then
identify necessity with that property.
My attempt to do this starts from the otherworldliness intuition. I argue
for a certain way of interpreting that intuition and of transforming it into an
approach to the question what necessity is. After that, I begin the quest for an
account of necessity anew, this time starting from the modalist intuition. I argue
that this intuition, too, can be transformed by natural steps into an account of
necessity. As it turns out, both maneuvers lead us to the same theory of necessity.
I take this to support my contention that this view can be regarded as capturing
both the otherworldliness and the modalist intuition.
Consider first the otherworldliness intuition, i.e. the intuition that necessary
propositions are those that are true in all scenarios that are alternatives to the way
things actually are. This intuition suggests that the ordinary-life practice that gives
rise to the concept of necessity is one in which we consider situations that we do
not believe to obtain, and ask ourselves what is true in them. There is more than
one common cognitive procedure that fits this description. In section 2, I argue
that the one in which modal concepts originate is our routine of representing to
ourselves certain scenarios and considering what would have been true if they had
obtained. We commonly express the outcome of such a thought experiment by a
counterfactual conditional. The core idea of my theory is that, roughly speaking,
for a proposition to be necessary is for it to play a certain distinctive role in the
truth-conditions of counterfactuals.
This view reverses the customary order of explanation. For the standard
view of counterfactuals, as propounded by Stalnaker and Lewis,5 analyzes them
in terms of the modal notion of a (metaphysically) possible world: if it had been
the case that p, then it would have been the case that q is (to simplify a bit) true
just in case q is true in all those possible worlds in which p is true and which are
otherwise as similar (close) to the actual world as is compatible with the truth
of p. I argue against this view in section 2.2. My discussion centers on a wellknown problem for the standard view: if the antecedent of a counterfactual is
metaphysically impossible, then there are no possible antecedent-worlds, so that
the standard account entails that the counterfactual is vacuously true. But it seems
very implausible to me that all counterfactuals with metaphysically impossible
antecedents are true. (It is metaphysically impossible for water to be an element,
but it does not seem true to say that if water were an element, everything would be
the case.) I adopt one of the obvious candidate solutions to this problem (which
has been developed in more detail by Daniel Nolan6 ): I reformulate the standard
account by simply replacing the concept of a possible world with the wider,
non-modal notion of a world. Worlds, which I think of as abstract entities of
some sort (possibly sets of propositions), comprise both possible and impossible
worlds. Impossible worlds are ordered by their closeness to the actual world, just
as possible worlds are. A counterfactual is true just in case the consequent is true
in the closest (possible or impossible) antecedent-worlds. The resulting theory of
counterfactuals does not use the concept of a possible world, and therefore leaves
us free to use counterfactuals to analyze necessity.
Such an analysis gains support from the observation that it seems natural to
paraphrase
P could not have failed to be true,
as
P would have been true no matter what (else had been the case),
or, equivalently, as: P is true and, for any situation whatsoever, if that situation
had obtained, P would still have been true. In section 2, I argue that, by qualifying
and refining this idea, we can develop an account of necessity. I also show that the
central idea of this account can be reformulated in terms of the familiar closeness
(similarity) account of counterfactuals: for a proposition to be necessary is for
it to be true in all worlds that have at least a certain degree of closeness to the
actual world.
Consider next the modalist intuition that the necessary truths are those
propositions whose truth is particularly secure or inexorable. This intuition
appears to rest on the idea that there is a dimension of inexorability on which
we can locate different truths. The necessary truths are those truths whose value
on that dimension is above a certain point. In section 3, I will argue that it is
this dimension of security or inexorability that we are talking about when we
ask ourselves, concerning some fact about our world, how easily it could have
failed to obtain. To say that a certain proposition has a high degree of security or
inexorability is simply to say that it could not easily have failed to be true. On the
account I will propound, how easily a state of affairs could have obtained depends
on the range of counterfactual situations in which it does obtain. If it obtains in
situations that depart only minimally from the actual world, then we are inclined
to say that it could easily have obtained. (Suppose that your favorite soccer team
would have won their last game if the goalkeeper had stood just an inch further to
the right in the fiftieth minute of the game. This gives you reasons for saying that
your team could easily have won.) If a state of affairs obtains only in situations
that depart very much from actuality, then we will instead say that it could not
easily have obtained. How easily a state of affairs could have obtained is thus
determined by the degree of closeness between the actual world and the closest
world in which the state of affairs does obtain. The degree of inexorability of a
true proposition is accordingly measured by the distance from the actual world
to the closest worlds in which the proposition is false. If we combine this with the
thought that metaphysical necessity is simply a high degree of inexorability, we
arrive, once again, at the conclusion that for a proposition to be metaphysically
necessary is for it to be true in every world that has at least a certain degree of
closeness to the actual world.
In section 3.3 I argue that this approach to metaphysical necessity can be
generalized to other kinds of necessity, such as nomic and conceptual necessity. I
think that these other kinds of necessity mark off degrees on the same dimension
of inexorability as metaphysical necessity, although they mark off different
degrees. The degree of inexorability that a proposition needs to have in order
to be conceptually necessary is higher, that which it needs to have in order to
be nomically necessary is lower, than that which is required for metaphysical
necessity. In section 4, I use these ideas to formulate formal definitions of
conceptual, metaphysical and nomic necessity.
My discussion of the modalist intuition yields the result that modal
propertiesnecessity and possibilitycome in degrees. I argue that, when we
talk about how easily a certain state of affairs could have obtained, we are talking
about its degree of possibility. The relation of comparative closeness of worlds,
too, can be interpreted as a relation of comparative possibility: to say that a
certain world is close to the actual world is to say that it could easily have been
actualized, i.e. that it has a high degree of possibility.7 The closeness relation is
therefore itself a modal relation. The definition of necessity in terms of closeness
that I give in section 4 reduces one modal concept to another. In order to reduce
the modal to the non-modal, it is still necessary to give a non-modal account
of the closeness relation. I tackle this task in section 6.
I find it plausible that modal facts must be grounded in non-modal facts. If
a proposition is necessary, then this modal status must ultimately be grounded
in its non-modal properties. We can sharpen this idea by drawing on the
modalist intuition that to be necessary is to be true in an especially secure
1.3. Resources
As a preliminary to carrying out the project I have outlined, I will give a
brief overview of the concepts and presuppositions on which my account rests.
Firstly, I will use the concepts of a things essence or nature, and of its
essential properties. The essence or nature of an object is what it is to be that
object. The essence of propane, e.g., is to be C 3 H 8 , since to be propane just
is to be C 3 H 8 . The essential properties of an entity are those that are part of
what it is to be that entity, i.e. those that make it the entity it is. It is, e.g., an
essential feature of propane that it is a compound of hydrogen and carbon; being
composed of these elements is part of what it is to be propane. By contrast, it is
merely accidental to propane that it is used for cooking.
It has long been common to define the notions of essence and of an essential
property in modal terms.10 (On this account, a property of a thing is essential to
it just in case the thing cannot exist and fail to have the property. The essence of
an object is a property that it cannot fail to have (if it exists) and that no other
thing could have had.) But recently Kit Fine11 argued that this characterization of
essence cannot adequately capture the underlying intuitive idea. As Fine points
out, it is a necessary feature of the number 2 to be a member of the set {2}, and
a necessary property of {2} to have 2 as a member. But while having 2 as an
element is part of what it is to be {2}, being a member of {2} is not part of what
it is to be 2. If Fine is right, as I think that he is, then there is no apparent reason
for thinking that the concepts of essence and essentiality need to be explained in
modal terms. The two notions can therefore be used in an account of necessity
without obvious threat of circularity. That is what I will do.
The way in which I conceive of my project is intimately bound up with
the concept of essence. I take the otherworldliness and modalist intuitions to
be intuitions about the essence of necessity. They amount, respectively, to the
intuition that truth in all alternatives to the way things are is the essence of
necessity, and that inexorability is the essence of necessity. And I understand my
present task as that of specifying the essence of metaphysical necessity, i.e. of
giving what is sometimes called a real definition of the property of metaphysical
necessity.
Secondly, I will avail myself of a conception of propositions as entities with
syntactic (sentence-like) structure. This conception allows us to understand the
notion of (narrowly) logical truth for propositions in a very intuitive way: a
logical truth is a proposition that is true in virtue of its logical structure alone.
This concept of logical truth seems to me to be non-modal, and the standard
(model-theoretic) way of spelling out its details makes no use of any modal
locutions.12
Thirdly, I will use the notion of a conceptual or analytic truth. I think that
this concept, too, can be understood non-modally: a proposition is an analytic
or conceptual truth just in case it is true in virtue of the fact that it is built
up from certain concepts in a certain way, and in virtue of the natures of these
concepts.
Given that the notions of conceptual and narrowly logical truth (and
the concomitant concepts of logical consequence, logical consistency, and so
forth) are non-modal, they can be used in an account of necessity without
circularity.
We can use the notions of narrowly logical consistency and conceptual
truth to define the concepts of analytic consistency and analytic consequence: a
proposition is analytically consistent just in case it is narrowly logically consistent
with the set of all conceptual truths. P is an analytic consequence of a set S of
propositions just in case P is a narrowly logical consequence of the union of S with
the set of all conceptual truths. The concepts of analytic consistency and analytic
consequence are non-modal. I will make extensive use of them throughout this
paper.
(implicitly and explicitly respectively) both the notion of a possible world and
the concept of a possible proposition.
Let me consider these worries one by one. I will first turn to the question
whether counterfactuals are to be analyzed in terms of the notion of a possible
world. The problematic occurrence of the term possible proposition on the
right-hand side of (2) will be the subject of the next section.
The basic idea underlying the standard account of counterfactuals has been
nicely stated by David Lewis in the opening sentence of his book on the matter:
If kangaroos had no tails, they would topple over seems to me to mean something
like this: in any possible state of affairs in which kangaroos have no tails, and
which resembles our actual state of affairs as much as kangaroos having no tails
permits it to, the kangaroos topple over.16
More formally, the counterfactual connective (for which I will use the symbol
!) is defined along the following lines:
(4) P ! Q is true just in case Q is true in all the closest (i.e. most similar)
metaphysically possible P-worlds.17
The possible-worlds account faces a well-known problem: If P is a metaphysically impossible proposition, then there are no possible worlds in which P
is true. In that case, it is vacuously true that Q holds in all the closest possible
P-worlds. The possible-worlds account therefore entails that, for any metaphysically impossible proposition P, all counterfactuals P ! Q are true. But we
have already seen in section 1.2 and in the last section that this is counterintuitive.
One strategy for avoiding this problem is to formulate the closeness account,
not in terms of the concept of a possible world, but in terms of the wider
(and non-modal) notion of a world, which covers impossible worldsworlds
in which impossible propositions are trueas well as possible worlds. Daniel
Nolan, among others, has argued in detail for this solution to the problem.18
The basic idea is simple: all worlds, not just the possible ones, are ordered with
respect to their closeness to the actual world. We can restate the closeness account
as follows:
(5) P ! Q is true just in case Q is true in all the closest P-worlds.
(5) is just like (4), except that the modal notion of a possible world has been
replaced by the wider concept of a world. For any impossible proposition P,
there are impossible worlds in which P is true. P ! Q is true if Q is true in
all the closest impossible P-worlds; it is false otherwise.
I believe that the foregoing considerations show that there are reasons quite
independent of the present project for replacing the standard account (4) by (5),
which makes no use of the notion of a possible world.
contingent (could
have been otherwise)
inexorability scale
(7) They could so easily have won the game. If the goalkeeper had stood
just an inch closer to the goalpost, the other team would not have scored
their second goal.
In general, if the state of affairs obtains in some worlds that depart only minimally
from the actual situation (as in the case of (7)), then we want to say that it could
easily have obtained. If it obtains only in worlds that depart very much from
actuality, then we say that it could not easily have obtained:
A: They could easily have won the game.
B: I dont think so. I think that they would have won only if Fred and Susie
had signed up for the team, Martha hadnt had a sore foot, Bugsy had
been sober, the weather had been nice, . . . .
This suggests that how easily a proposition P could have been true is a matter
of how much the worlds in which P is true depart from actuality. The closer the
closest P-worlds are, the more easily P could have been true.23
It seems plausible to me that could easily have been the case is only
one among a whole range of ordinary-language expressions that can be used
to express degrees on this dimension. Other expressions of this kind arguably
include:
It almost happened.
It was within a hairs breadth.
This was a close call.
This was a narrow escape.24
And so forth. For example:
He was almost killed. If he had stood five inches further to the right, the
brick would have smashed his head.
There is a counterfactual situation that departs only minimally from ours, such
that he would have been killed if that situation had obtained. This supports the
claim that he was almost killed.
I suggest that could have been otherwise and could not have been otherwise
express degrees on the very same dimension that we are talking about when we
ask how easily things could have been otherwise. This idea seems very plausible.
Consider:
Freds house could easily have been destroyed by yesterdays earthquake. (If
the earthquake had been just a little bit stronger, the house would have been
destroyed.)
Susies house could not as easily have been destroyed. (The earthquake would
have had to be much stronger to destroy Susies house.)
Bugsys house could not possibly have been destroyed. (Earthquakes of
sufficient strength never occur in this area.)
This sequence of sentences feels intuitively like a progression along a single
dimension.
This suggests that, when we are speaking about a propositions degree of
inexorability, we are talking about how easily it could have been false. To say
that it has a high degree of inexorability is to say that it could not easily have
been false, i.e. that the closest worlds where it is false are far away from the
actual world. In the example of the last paragraph, e.g., the fact that Freds
house survived the earthquake has a low degree of inexorability, the fact that
Susies house survived has a somewhat higher degree, the fact that Bugsys house
survived has the highest degree of the three facts.
Which degree on the inexorability scale is expressed by could not have been
otherwise depends on the context. In philosophical discussions, the phrase often
expresses metaphysical necessity. In ordinary life, it often expresses a lower degree
of inexorability. (Seeing that Susie was head and shoulders above the competition,
I can truly say that she could not have failed to win the tournament. This does not
mean that her victory was metaphysically necessary.) It is an interesting question
how it is determined which degree of inexorability is expressed by the phrase in
any given context.
The key to the answer may lie in a conspicuous feature of our use of could:
When we say that neither A nor B could possibly have failed to be true, we
cannot add, in the same breath, that A could more easily have failed to be true
than B. We cannot say, Neither Smith nor Jones could possibly have won the
tournament, and Smith could have done so more easily than Jones. By saying
that A and B could not have failed to be true, we take them out of the range of
propositions for which we can raise the question of how easily they could have
failed to be true. In a different context, it may be true to say that A and B could
have failed to be true, and in that context, it makes perfect sense to say that A
could more easily have failed to be true than B. For instance, when we move into
a context in which could expresses metaphysical possibility, it becomes true to
say that both Smith and Jones could have won the tournament. And if Smith is
much better than Jones, then we can say in this new context that Smith could
more easily have won than Jones.
The following analogy illuminates my attempt to make sense of these data.
Suppose that I want to buy a car. I am interested in the differences between the
prices of different models. But only within certain limits. I do not much care how
the price of a Rolls Royce compares to that of a Mercedes, since I know that I
cannot afford either car. There is a certain maximum amount of money I can
spend, and differences between prices that are higher than that are of no interest
to me. That is how it often is when we are comparing the values of different
objects along a certain dimension, and I claim that it is also how it is when
we are comparing the values of different truths on the inexorability scale. For
the purposes at hand, we may be interested in the differences between degrees
of inexorability below a certain point on the scale, but we may not care about
the differences between the different values that are above that point. I would
like to suggest that could not have been otherwise marks what we may call the
indifference point on the inexorability scale, i.e. that point on the scale above
which we do not wish to distinguish degrees of inexorability in the given context:
when it is true to say, in a given context, that neither A nor B could possibly
have been the case, then we cannot, in the same context, ask which of A and B
could more easily have been the case. The location of the indifference point on
the inexorability scale can vary from context to context.
I suggest that the use of could not have been otherwise to express metaphysical necessity is simply a special case of this. The contexts in which the
phrase expresses the notion of metaphysical necessity are simply those in which
the indifference point has a specific (very high) value on the inexorability scale.25
On this view, for a proposition to be metaphysically necessary is for it to be true
in every world that is at most a certain distance away from the actual world. This
is just the idea that we arrived at in section 2, as a result of trying to develop
the otherworldliness intuition. We have thus gained independent support for the
view via two different routes.
the closest world where w obtains. Since the closest world in which w obtains is
w itself, this amounts to saying that a worlds degree of possibility is simply its
degree of closeness. The relation of comparative closeness can thus naturally be
thought of as a relation of comparative possibility of worlds. To ask how close
a world is to the actual world is to ask how easily the world could have been
actualized, what degree of possibility it possesses.
In the last section, I analyzed inexorability in terms of closeness. This account
amounts to a definition of comparative necessity for propositions in terms of
comparative possibility for worlds. It reduces one modal notion to a more basic
modal concept. In order to reduce the modal to the non-modal, we still need
to give a non-modal account of the closeness relation. This will be the task of
section 6. Drawing on the results of the previous sections and on those of my
(2006), I will define closeness in terms of concepts which, in my opinion, are not
themselves to be analyzed in modal terms, such as the concepts of explanation
and of a law of nature.
The picture I sketched is summarized in the table below (an arrow means is
definable in terms of ):
Comparative necessity (aka comparative inexorability) of
propositions
Modal concepts
Comparative possibility (aka comparative closeness) of worlds
Non-modal
concepts
In the next section, I will attempt to provide further evidence for the claim that
necessity comes in degrees, by arguing that this view allows us to give a unified
account of different kinds of necessity.
to be metaphysically necessary, and the latter degree is, in turn, higher than that
which is required in order for a proposition to be nomically necessary.
On this view, there are points on the inexorability scale, v n , v m , v c , such that
to be nomically necessary is to have some value above v n on the inexorability
scale, to be metaphysically necessary is to have some value above v m , and to be
conceptually necessary is to have some value above the still higher point v c .
Inexorability / necessity
scale
conceptually
necessary props.
vc
vm
metaphysically
necessary props.
nomically necessary
vn
@
when they were presented with counterfactuals whose antecedents are analytic
falsehoods, such as If there were triangles that did not have three angles, . . . .
Such counterfactuals are quite different from the likes of If Thatcher were my
mother . . . , inasmuch as their antecedents contradict themselves. Some of my
subjects felt that this made it hard for them to understand what scenario they
were supposed to envisage and reason about hypothetically, and therefore made
it difficult for them to assign truth-values to the counterfactuals in a non-trivial
way. Other subjects saw no such difficulty. I am undecided.
I said in section 2.2 that we can permit counterfactuals with metaphysically
impossible antecedents to have non-trivial truth-conditions if we let impossible
worlds figure in the theory of counterfactuals alongside possible worlds. A
metaphysically impossible world can be analytically consistent (i.e., the propositions true at the world may be jointly narrowly logically consistent with all
conceptual truths). Whether or not we want to take the further step and allow
analytically inconsistent worlds to figure in our theory depends on whether we
want counterfactuals with analytically false antecedents to have non-trivial truthconditions. If we do not, then we will have no need for analytically inconsistent
worlds. We can then define worlds as maximal analytically consistent sets of
propositions. If we want counterfactuals with analytically false antecedents to
have non-trivial truth-conditions, then we need to make room for analytically
inconsistent worlds. We can do so by defining worlds, not as maximal analytically
consistent sets of propositions, but simply as maximal sets of propositions,
without stipulating that they need to be analytically consistent.
For our present purposes, we need not decide between the two views. We
merely need to argue that on either view, the analytically consistent worlds form
a sphere around the actual world, i.e.
(8) Analytically consistent worlds are closer to the actual world than analytically inconsistent worlds.
If there are no analytically inconsistent worlds, then (8) is vacuously true. But I
think that (8) is also true if there are analytically inconsistent worlds. In order to
understand what motivates this assumption, note that, as long as an assumption
we hypothetically entertain is analytically consistent, no analytic inconsistency
should emerge in the course of our reflection on what would have been the case
if the assumption had been true. More precisely, where A is any analytically
consistent proposition, every proposition C for which A ! C is true is
analytically consistent with A. This finding can be explained by (8), as is easy
to see: Suppose that A is an analytically consistent proposition, so that there
are analytically consistent worlds in which A is true.31 If (8) is true, then the
analytically consistent A-worlds must be the closest of all A-worlds. And since
the propositions true at these worlds must be mutually analytically consistent,
there can be no proposition C that is analytically inconsistent with A such that
A ! C. (8) seems to me to provide the best explanation for the fact that
no analytic inconsistency can emerge in hypothetical reasoning from an analytically consistent assumption, and it is strongly supported by this explanatory
power.
Since the analytically consistent worlds are all and only the conceptually
possible worlds, we can infer from (8) that the conceptually possible worlds form
a sphere around the actual world.
Nomic necessity. I characterized nomic necessity as a kind of necessity that is,
in some way, associated with the laws of nature (as well as with the metaphysical
necessities). But what exactly is the association between nomic necessity and the
laws? There is no consensus on the matter. According to some philosophers,
nomic necessity is a kind of necessity that attaches to the laws of nature, and to
all the propositions that are metaphysically necessitated by the laws. On another
view, nomic necessity also attaches to the truths about which principles are laws,
e.g. to true propositions of the form P is a law and P is not a law.32 On yet
another account, truths about which principles are laws are nomically necessary,
whereas the laws themselves are not.
My sympathies are with the third of these views. Let me explain why. Analytic
philosophers often (though by no means universally) assume that a universal
generalization cannot be a law of nature unless it is true without exception. But
it seems to me that, inasmuch as we are in the business of trying to capture the pretheoretical notion of lawhood, this assumption ought to be controversial. That
is, it does not seem obvious to me that there is anything in the folk concept of a
law that precludes the existence of exceptions to a law. (For many centuries belief
in miracles was very common as a central component of popular religious faith,
and on one natural and common way of understanding the notion of a miracle,
it involves a violation of natural law.) Moreover, I think that there are theoretical
pressures that make the view that laws can have small exceptions attractive. (I
say more about this issue in my (2006). See Lange (2000) for a considerably more
detailed discussion.) On this view, a universally quantified proposition can be a
law, even though it is not, strictly speaking, true. But a false proposition cannot
have any kind of necessity. I therefore do not subscribe to the idea that there
is any kind of necessity that attaches to all laws of nature. Instead, I think that
nomic necessity attaches to the truths about which principles are laws of nature,
as well as to all propositions that are metaphysically necessitated by the truths
about which principle are laws.
On this view, the nomically possible worlds are just those metaphysically
possible worlds that have the same laws of nature as the actual world. It is
controversial whether these worlds form a sphere around the actual world, but I
have argued in detail in my (2006) that this is true.33
The discussion of this section suggests an answer to the question of which
commonality between conceptual, metaphysical and nomic necessity makes them
species of a common genus: for each of the three properties, there is a point on
the inexorability scale such that for a proposition to have the property is for it
to have a degree of inexorability above that point. There are many other species
of the same genus. For every sphere, there is a kind of necessity that attaches to
all and only those propositions that are true in every world in that sphere. In
addition to the kinds of necessity that have familiar names, such as conceptual,
metaphysical and nomic necessity, there are innumerable other, nameless kinds
of necessity. For any world w, there is a sphere that contains all and only those
worlds that are at least as close to the actual world as w, and there is a kind of
necessity that attaches to all and only those propositions that are true in every
world in that sphere. This kind of necessity may not have a familiar name in
English, although we can describe it. We can use the phrase the negation of P
could less easily have obtained than w.
epistemically necessary is for its truth to be particularly secure in some sense; but
the security is epistemic, not ontic or metaphysical. The security of the proposition
consists in the fact that its truth is guaranteed by something we know, or that
we have an a priori guarantee that the proposition is true, or something of that
kind. There is no objective metaphysical security that is distinctive of the epistemic
necessities.
I am equally doubtful that any of the other properties listed under (9) capture
degrees of metaphysical inexorability (in the sense that having that property
consists simply in being located above a certain point on the inexorability scale).
Consider the case of biological necessity. If biological necessity captures a degree
of metaphysical inexorability, then the truth of every biological necessity must be
more secure and inexorable than that of any proposition that is not biologically
necessary. But this seems intuitively implausible. There are laws of physics and
chemistry that are not biological necessities (since they are not laws of biology),
but it seems implausible that their truth is less inexorable than that of biological
necessities. Similarly, I am not convinced that mathematical necessity captures
a degree of inexorability. It is mathematically necessary that 2 exists; it is not
mathematically necessary that Aristotle is not a musical performance (since this
is not a truth of mathematics).34 But the truth of the latter proposition does
not seem less inexorable than that of the former. (The obstacles that prevent
Aristotle from being a musical performance do not seem less weighty than those
that prevent 2 from failing to exist.)
Although the notions of biological and mathematical necessity do not mark
off degrees of inexorability, they may be definable in terms of notions that do.
Two ways of defining new modal concepts from notions of necessity deserve
attention in this context. Following Kit Fine, I will call them restriction and
relativization.35 To say that a notion N is defined from a kind of necessity
N by restriction is to say that a propositions falling under N consists in the
combination of two things: (i) The fact that it possesses the kind of necessity N ;
and (ii) the fact that it meets certain additional conditions. To use an example of
Fines,36 there is some plausibility to the idea that the concept of mathematical
necessity is defined from the notion of metaphysical necessity by restriction. I find
it attractive to say that a propositions being mathematically necessary consists
in its being metaphysically necessary for a certain reason, viz. the reason that
it is a mathematical truth. To say that a notion N is defined from a kind of
necessity N by relativization is to say that a propositions falling under concept
N consists in its being necessitated, relative to the kind of necessity N , by certain
propositions P. For example, it has some plausibility to say that the concept of
biological necessity is defined from the concept of metaphysical or conceptual
necessity by relativization to the laws of biology: a propositions being biologically
necessary consists in its being metaphysically or conceptually necessitated by the
laws of biology.37
In the remaining sections I will be concerned exclusively with those
modal notions that ascribe degrees of metaphysical inexorability, as conceptual,
metaphysical and nomic necessity do, whereas the concepts listed in (9) do not.
Whenever I speak of kinds of necessity in what follows, I will have only those
properties in mind.
I believe that modal classifications are prior to, and more basic than, the concept
of closeness.
I: What, then, do you make of the perfect correlation between the two
relations? How do you explain it? If we identify the two relations, as I propose
we do, then this question does not arise. But if we resist the identification, then
we need some explanation.
Albert: Consider again the fact that the metaphysically possible worlds are
closer than the metaphysically impossible worlds. I think that it is the special
modal status of the metaphysically possible worlds that explains their position
in the closeness ordering: the metaphysically possible worlds are closer than
other worlds because they are metaphysically possible. Now, the fact that the
metaphysically possible worlds are closer than the metaphysically impossible
worlds is just a special case of the general correlation between comparative
possibility and comparative closeness. We should expect to find a unified explanation of this correlation. Hence, if we say that the special position of the
metaphysically possible worlds in the closeness ordering is due to the fact that
they are metaphysically possible, then we should say that it is true quite generally
that a worlds degree of possibility determines and explains its position in the
closeness ordering. The closeness ordering, in turn, determines the truth-values
of counterfactuals in the familiar way: a counterfactual is true just in case its
consequent is true in the closest antecedent-worlds.
I: This account strikes me as unnecessarily complicated. If the relations of
comparative closeness and comparative possibility are perfectly correlated, then
we do not need both of them in the theory of counterfactuals, and it seems to be
a demand of good methodology to get rid of one of them. Instead of explaining
the closeness ordering by appeal to comparative possibility and then explaining
the truth-conditions of counterfactuals in terms of closeness, you could as well
directly define the counterfactual connective in terms of comparative possibility.
You could simply say that P ! Q is true just in case Q is true in those
P-worlds that have the highest degree of possibility.
Albert: So much the better. This move only makes my theory simpler and
more attractive. I can now do without the relation of closeness altogether. Instead,
I am using the notion of comparative possibility, which plays exactly the same
role in my theory as the concept of closeness does in the standard account of
counterfactuals.
I: I deny that you have really gotten rid of the closeness relation. The meaning
of the term closeness as used in the literature on counterfactuals (or at least as
I am using it) is determined by its theoretical role: closeness refers to whatever
relation induces the ordering of worlds that figures in the correct theory of
counterfactuals. If it is the relation of comparative possibility that induces this
ordering (as on your account), then the term closeness refers to the relation
of comparative possibility. Hence, the relations of comparative closeness and
comparative possibility are not distinct after all. They are one and the same
relation.
4. Defining Necessity
According to the results of the previous section, conceptual, metaphysical
and nomic necessity are different species of a common genus: each of them is a
propositions property of having at least a certain degree of inexorability. Each
of the three properties should therefore have a definition of the form:
(11) To be metaphysically (conceptually, nomically) necessary is to be true
throughout that sphere of worlds around the actual world that meets
condition C M (C C , C N ),38
where the term actual is, once again, to be understood in a non-rigid sense.
In order to complete our quest for a definition of metaphysical (conceptual,
nomic) necessity, we still need to find a suitable replacement for C M (C C ,
C N ). We must replace this dummy by the statement of some condition which,
in every possible world, singles out the sphere of all and only the metaphysically
(conceptually, nomically) possible worlds. This section will be devoted to finding
a suitable condition.
My strategy for achieving this goal will be somewhat indirect. I mentioned in
the introduction that I find it plausible that the modal properties of a proposition
need to be grounded in its non-modal features. In section 4.1 I will propose a
tentative answer to the question of which non-modal features can ground the
conceptual, metaphysical and nomic necessity of a proposition. I think that a
correct definition of necessity needs to be compatible with that answer. In section
4.2, I will motivate a specific way of completing the definitional schema (11) by
arguing that it yields a definition of necessity that meets that constraint.
about the world. I will use this view as my working account of the features that
ground the modal status of metaphysically necessary propositions.
I do not claim that this working account is right in all its details. It is enough
for my purposes if it is approximately correct. It might be that the mathematical
truths do not owe their special modal status to the fact that they are truths
of mathematics, but to some more general feature, e.g. to the fact that they
are truths that are purely about abstract entities, and not even in part about
the concrete world. Such a view would merit exploration. For the purposes of
this paper, however, I will make do with the working account stated in the last
paragraph.
Given our view about what grounds metaphysical necessity, it is not hard
to come up with an account of what grounds nomic necessity. On my account,
the extension of nomic necessity contains all and only those propositions that
are metaphysically necessitated by the truths about which principles are laws of
nature. It is therefore natural to think that every nomic necessity owes its special
modal status to the fact that it flows from the mathematical and (conditionalized)
essential truths and the truths about which principles are laws.
The reader may worry that the mathematical truths, the conditionalized
essential truths and the truths about which principles are laws do not form an
interesting and unified class. Why should it be precisely these three categories of
truths that are capable of grounding modal force? In section 5, I will try to assuage
such worries by arguing that, if we endorse the view I stated in this section, we
have an attractive way of cashing out the idea that the necessary propositions are
those propositions whose truth is underwritten by particularly deep features of
the world order. The picture I will sketch in section 5 will make it understandable
why it is precisely the relevant three categories of truth that ground the special
modal status of the metaphysical and nomic necessities.
I also claim that the following two principles are part of the definition of the
concept of closeness:
(R M ) Analytically consistent worlds in which all actual mathematical and
conditionalized essential truths hold are closer to the actual world
than all other worlds.
(R N ) Analytically consistent worlds in which all actual mathematical and
conditionalized essential truths hold, and which have the same laws
of nature as the actual world, are closer to the actual world than all
other worlds.
(If we do not allow for analytically inconsistent worlds, then the words analytically consistent can be deleted from (R M ) and (R N ).) The expression actual
in (8), (R M ) and (R N ) is to be understood in the non-rigid sense. As I said
in section 1.3, I take the notion of analytic consistency to be non-modal, and
therefore believe that we can make (8), (R M ) and (R N ) parts of our definition
of closeness without thereby importing any modal content into the notion of
closeness.
The degree of closeness between two worlds is the resultant of weighting
different similarities and dissimilarities between them. A theory of closeness
that incorporates the rules (R M ) and (R N ) suggests the following picture of the
principles for weighting similarities. These principles single out three disjoint sets
of truths and order them by the weight that attaches to the truth of them at
an analytically consistent antecedent-world: firstly, there are mathematical and
conditionalized essential truths. If all actual mathematical and conditionalized
essential truths hold at an analytically consistent antecedent-world, then this
makes the world closer than any world that does not meet this condition. Secondly, there are truths about which principles are laws. If all actual mathematical
and conditionalized essential truths and all actual truths about the laws hold
at an analytically consistent world, then this world is closer than any world that
does not meet this condition. Thirdly, there are truths about the individual events
that take place in the natural world. These truths carry less weight than truths
of the two other kinds.
On the basis of the foregoing considerations, I suggest the following definitions of conceptual, metaphysical and nomic necessity:
To be conceptually necessary is to be true throughout the sphere around
the actual world that contains all and only the analytically consistent
worlds.
To be metaphysically necessary is to be true throughout the sphere around
the actual world that contains all and only those analytically consistent
worlds in which all actual mathematical and conditionalized essential truths
hold.
But what, on such a view, is the commonality between nomic necessity and
metaphysical necessity that makes them both kinds of necessity? Each of the
two kinds of necessity is the property of being an analytic consequence of a
certain set of propositions. But not every property of being a logical consequence
of a certain set of propositions is a kind of necessity. What distinguishes those
properties of this kind that are kinds of necessity from those that are not? The
account envisaged leaves us completely in the dark about the answer.
It is one of the most fundamental facts about metaphysical necessity that
it is a kind of necessity. Hence, whatever else is true of metaphysical necessity,
it must have the defining feature of the higher-order property of being a kind
of necessity. An account of metaphysical necessity that makes it a mystery what
this feature might be thereby lays itself open to the suspicion that it has failed to
capture that feature of metaphysical necessity which makes it a kind of necessity;
that it has, in fact, failed to capture that which makes it a modal property. But a
theory of metaphysical necessity that does not capture the fact that it is a modal
property must be fundamentally flawed.
I suspect that the account (12) will also be incapable of accommodating the
modalist intuition, and will thereby fail to achieve one of the central goals of a
metaphysical account of modality. According to the modalist intuition, it is the
distinctive feature of the metaphysical necessities that their truth is particularly
secure and inexorable. There must be some dimension of security or inexorability,
such that for a proposition to be necessary is simply for it to have a particularly
high value on that dimension. A good account of metaphysical necessity should
tell us what this dimension is. And it should offer as the defining feature of
metaphysical necessity the property of having a high value on that dimension.
The account I offered in this paper meets this condition: The inexorability of
a proposition is measured by the distance from the actual world to the closest
worlds in which it fails to be true, and necessity is defined as the property of
having a high value on that scale. But what about (12)? There is no mention of
the inexorability scale in this proposed definition of necessity.
The failure to accommodate the modalist intuition is ultimately the same as
the failure to explain what makes metaphysical necessity a kind of necessity. For
I argued in section 3.3 that what makes a property a kind of necessity is that it
is the property of having at least a certain degree of inexorability. The modalist
intuition therefore in effect amounts to the intuition that metaphysical necessity
is a kind of necessity.
I expect the proponent of (12) to reply as follows: I can see no reason
why I should not endorse the view outlined in section 3.1 about the dimension
of inexorability that is relevant: the inexorability of a truth is measured by the
distance from the actual world to the closest worlds in which the proposition
is false. It is true that this dimension of inexorability is not mentioned in my
definition of the concept of metaphysical necessity. But there is nonetheless a clear
sense in which, even on my account, the concept of metaphysical necessity singles
out a segment of the inexorability scale; for it singles out just those propositions
whose position on the inexorability scale is above a certain point. By the same
token, I can give an account of what makes a property a kind of necessity. What
the different kinds of necessity have in common is that each singles out the set
of just those propositions that are true in all worlds that are at most a certain
distance away from the actual world. And I can say that that is what makes those
different properties kinds of necessity.
Such an approach to modality would be rather close to my own. In particular,
its proponent would have foregone the one advantage that account (12) seemed to
have over my theory, namely its greater simplicity. It is true that the definition of
metaphysical necessity proffered in section 4.2 mentions spheres, whereas there is
no mention of them in (12). But if we combine (12) with the account mentioned
in the last paragraph of what makes a property a kind of necessity, then we have
not, after all, gotten rid of the story about closeness and spheres, with all its
complexity. We have just relocated it: instead of making it a part of the definition
of metaphysical necessity, we have only made it part of our account of what
makes metaphysical necessity a kind of necessity. Our overall theory of modality
has not become simpler.
Simplicity considerations are therefore not relevant to a decision between
the two accounts. Instead, we need to ask whether it is better to write the
connection between necessity and closeness into the definition of metaphysical
necessity, or to mention it only in the account of what makes metaphysical
necessity a kind of necessity. I think that the former option is better. Let me
briefly say why. The proponent of the rival account believes that the defining
feature of the property of being a kind of necessity is the property of singling
out those propositions that have a specific location on the inexorability scale.
But this location on the inexorability scale is not mentioned in her definition
of metaphysical necessity. This means that the fact that metaphysical necessity
is a kind of necessity is not reflected in her specification of what metaphysical
necessity is, in her proposed statement of the essence of metaphysical necessity.
On the rival account, it is therefore not an essential, but merely an accidental,
feature of metaphysical necessity that it is a kind of necessity. On the rival
account, one could know what metaphysical necessity is, and also know what
it is for a property to be a kind of necessity, and yet be in no position to know
that metaphysical necessity is a kind of necessity. But this, I think, is rather
implausible.
By the same token, I also believe that the rival account, even when developed
in the way suggested in the fictional reply of its proponent, still fails to accommodate the modalist intuition. As I said in sections 1.2 and 1.3, the modalist intuition
is an intuition about what it is for a proposition to be necessary, about the essence
of necessity. According to the modalist intuition, it is the essence of necessity to
be the property of having a high degree of inexorability. Whatever property of
the necessary propositions constitutes their high degree of inexorability must be
the defining feature of necessity, and must figure as such in a specification of the
essence of necessity. But when we now consider (12), it is obvious that the feature
fact that L is a law explains the general fact that events conform to L
(i.e. the fact that L is a law explains why L is true).47
To take an example, consider
(Law of Gravitation) Any two bodies of masses m 1 and m 2 that are at
distance d of each other attract one another with a
force of strength Gm 1 m 2 /d 2 ,
where G is the gravitational constant. Assume that (Law of Gravitation) is a law
of nature. Last week, Mars took a certain path through space in accordance with
(Law of Gravitation). I believe that the fact that (Law of Gravitation) is a law is
one of the factors that together explain why Mars took the path it did. And I
also believe that the fact that (Law of Gravitation) is a law explains the general
fact that events conform to (Law of Gravitation); that is, it explains why bodies
of masses m 1 and m 2 that are at distance d of each other attract one another
with a force of strength Gm 1 m 2 / d 2 .
(L) seems very plausible to me. Most other people I have asked find the
principle plausible, too, and this makes me hope that the reader will find it
reasonable as well. Unfortunately, any serious discussion of (L) would have to
take up a lot of space, and is therefore beyond the scope of this paper.48
Now suppose that (Law of Gravitation) is a fundamental law of our world,
i.e. that it is a law, and that this fact cannot be explained by appeal to other, more
fundamental laws. Consider,
(13) If (Law of Gravitation) had not been a law, then events would still have
at least approximately conformed to it.
No one I asked believed that this counterfactual was true.
How can we explain our unwillingness to accept (13)? I will focus on an
explanation with considerable intuitive appeal. The only reason why events
conform to (Law of Gravitation) in our world is that (Law of Gravitation) is
a law. But that reason is absent in an antecedent-world. Hence, even if the events
of an antecedent-world (approximately or even perfectly) conform to (Law of
Gravitation), their conformity to the law does not have the same explanation as
in our world. That is why it contributes nothing to closeness. Antecedent-worlds
that conform to (Law of Gravitation) are no closer than those that do not. This
is why there is no reason for accepting (13).
The foregoing considerations suggest the following generalization of (c):
(C) If some fact f obtains in both of two worlds, then this similarity
contributes to the closeness between the two worlds if and only if f
has the same explanation in the two worlds. (In the special case in which
Diagram 1
Diagram 2
worlds. According to the results of section 5.1, the closest no-g worlds are those
analytically consistent no-g worlds that maximize match in facts with the same
explanations as in our world. We can represent the situation diagrammatically:
g and facts that are explained (at least in part) by g are represented by asterisks
in Diagram 2, while all other facts are represented by dots. g cannot obtain in
any analytically consistent no-g world. And if one of the other facts represented
by asterisks obtains in a no-g world w, it cannot have the same explanations as
in our world, and the sharing of the fact is therefore irrelevant to the closeness
between w and our world. Only match in facts represented by dots (dot facts,
for short) can contribute to closeness. The closest no-g worlds will be those
analytically consistent worlds that maximize match in dot facts that have the
same explanations as in our world. If there are analytically consistent no-g worlds
that share all the dot facts with our world, then these will be the closest no-g
worlds.
Of course, there may be no such worlds. It could be that the dot facts of
Diagram 2the facts that are not g and are not explained by gtogether entail
g, so that no analytically consistent no-g world can contain all the dot facts. There
are many cases of this kind. Suppose, e.g. that determinism is true,51 and that
causation always proceeds forward in time. Let x be the fact that a specific kind
of event occurred at time t. Given the temporal asymmetry of causation, x does
not contribute to explaining any events before t. x also does not contribute to
explaining any laws of nature (for laws of nature are not explained by individual
events).52 But under determinism the facts about the worlds history before t,
together with the laws of nature, entail that x obtains.53 Hence, either the history
before t is different in the closest no-x worlds, or some actual law is violated
in these worlds (and thus fails to be true), or both.54 To take another example,
suppose that L 0 is a fundamental law of nature, and that L 1 is a derived law
whose lawhood is explained by the lawhood of L 0 and by certain mathematical
truths that are used in the derivation. The fact that L 0 is a law and the relevant
mathematical truths jointly entail that L 1 is a law. Thus, there are facts that are
distinct from the fact that L 1 is a law and are not explained by the lawhood of
L 1 , and which jointly entail that L 1 is a law, and some of these facts must be
absent in the closest worlds in which L 1 fails to be a law.
If the dot facts of Diagram 2 entail g, then in every analytically consistent
no-g world some of the dot facts are absent. But the closest no-g worlds are those
that maximize match in dot facts with the same explanations. Hence, they contain
most of the dot facts of Diagram 2. Let us say, for the sake of definiteness, that
h is absent in the closest no-g worlds (which means that match with respect to
facts that are actually explained by h cannot contribute to the closeness of these
worlds, so that the facts that are actually explained by h might also be absent
in the closest no-g worlds), but that all dot facts except h and the facts actually
explained by h are present in the closest no-g worlds. Thus, the facts represented
by dots in Diagram 3 are all and only those actual facts that obtain in the closest
no-g worlds and have the same explanations as in our world:
Diagram 3
Now ask yourself what the closest worlds in which f fails to obtain look
like. Once again, the closest no-f worlds will be those analytically consistent no-f
worlds that maximize match in facts with the same explanations as in our world.
The asterisks in Diagram 4 represent f and the facts explained (at least in part)
by f . All other facts are represented by dots.
Diagram 4
f cannot obtain in any analytically consistent no-f world, and if other asterisk
facts of Diagram 4 obtain in an analytically consistent no-f world w, they
cannot have the same explanations as in our world, and the match in those
facts is therefore irrelevant to the closeness between w and our world. The only
similarities between an analytically consistent no-f world and the actual world
that can be relevant to closeness are similarities in the dot facts of Diagram 4.
But even if the dot facts of Diagram 4 are jointly analytically consistent with the
non-obtaining of f , so that there is an analytically consistent no-f world that
shares all the dot facts with our world, the combined weight of these similarities
is smaller than the combined weight of match in the facts represented by dots
in Diagram 3, simply because the dot facts of Diagram 4 are a proper subset of
those of Diagram 3. Hence, the closest no-f worlds are less close than the closest
no-g worlds. The difference in closeness is even more striking if the dot facts of
Diagram 4 are jointly analytically inconsistent with the non-obtaining of f , so
that there are no analytically consistent no-f worlds that share all of these facts
with our world.
The upshot of the foregoing considerations (which were not rigorous, and
were merely intended to convey the intuitive picture) is the following rule of
thumb:
(D) By and large, if a fact f is explanatorily prior to a fact g, in the sense that
f contributes to explaining g but not vice versa,55 then the closest no-f
worlds are farther away than the closest no-g worlds; that is, f could
less easily have failed to obtain than g.56
For the most part, if f is represented by a node closer to the bottom of Diagram 1
than g, then the truth that f obtains is more secure and inexorable than the truth
that g obtains. By and large, the more explanatorily basic and fundamental a fact
is (the greater the range of facts which it underlies and explains), the greater its
degree of necessity.57
events is exactly parallel: while facts about which principles are laws contribute
to explaining facts about individual events (in accordance with principle (L) of
section 5.1), facts about individual events do not explain facts about lawhood. (It
it important not to misunderstand these theses of explanatory asymmetry. I do
not, of course, claim that every mathematical fact figures in the explanation of
every truth about what the laws are, or even that every mathematical fact figures
in the explanation of some members of the category of truths about the laws.
My thesis is that there is an explanatory asymmetry between the two categories
or classes of propositions, not that there is an explanatory asymmetry between
every individual mathematical truth and every individual truth about lawhood.
The explanatory asymmetry between the categories consists in the fact that some
members of the category of mathematical truths figure in the explanation of some
members of the category of truths about the laws, whereas no truths about the laws
figure in the explanation of any mathematical truths. Similarly for the relationship
between the class of truths about lawhood and the class of truths about individual
events.)
The foregoing considerations suggest the following picture: There is a
classification of facts that carves nature at its joints, that corresponds to an
important metaphysical distinction. On the one side of it, there are the facts that
are, at least in part, about the realm of concrete entities, on the other side there
lie the facts of pure mathematics (and perhaps other facts as wellperhaps all
facts that are purely about abstract entities). Facts of the first kind are further
subdivided into those about which laws govern the realm of concrete objects,
and those about the individual events that occur in the concrete realm. These
three major metaphysical categories of facts are ordered with respect to how
explanatorily fundamental they are: the class of mathematical facts (or: of all facts
about pure abstract entities?) is explanatorily prior to, or more fundamental than,
the class of facts about concrete entities. And among the facts about concrete
entities, those about the laws that govern the concrete realm are explanatorily
more fundamental than those about individual events. This ordering of the three
classes of facts by their degree of explanatory fundamentality is represented in
Facts about
Facts about
individual events
Mathematical facts
Diagram 5
Facts about
concreta
the diagram. The more fundamental facts are lower in the diagram, and arrows
indicate explanatory relations in the usual way.
Our rules for determining the closeness ordering of worlds follow the contours of these three fundamental metaphysical categories of facts. In accordance
with principle (D), they determine that the closest worlds in which the actual
truths of mathematics fail to hold are less close than the closest worlds with
different laws of nature, and that the latter are, in turn, less close than worlds
that differ from ours only with respect to the individual events that occur in the
natural world.
Let us look at the same issue from a slightly different angle. Suppose that
you want to know what would have been the case if a certain fact x had not
obtained. (I will once again assume that it is analytically consistent to say that x
fails to obtain, so that (according to (8)) the closest no-x worlds are analytically
consistent worlds.) You need to find out what the closest no-x worlds are like.
Now, given that x actually obtains, the totality of all facts is obviously analytically
inconsistent with the non-obtaining of x. Hence, not all actual facts can obtain
in the closest no-x worlds. The question is which of the actual facts obtain in
these worlds.
How can we find the answer to this question? Think of it as a game. You
are given Diagram 1, and you need to remove enough nodes from the diagram
to ensure that the facts represented by the remaining nodes are analytically
consistent with the assumption that x does not obtain. Whenever you remove
one of the nodes, n, from the diagram, every node to which a chain of arrows
leads from n also disappears. Each of the nodes is assigned a certain number,
which measures the weight with which the sharing of this fact contributes to the
closeness between two worlds if the fact has the same explanation in the two
worlds. You remove nodes from the diagram until you reach a point at which the
remaining nodes represent facts that are jointly analytically consistent with the
non-obtaining of x. At this point, your score will equal the sum of the numbers
assigned to the nodes that remain in the diagram. The goal of the game is to
maximize this final score. Each optimal way of playing the game (each way of
playing the game that maximizes the final score) corresponds to a set of closest
no-x worlds, namely those analytically consistent no-x worlds that share with
our world all the facts represented by nodes that remain in the diagram at the
end of the game.
It is important for our purposes to take into account one important rule of
strategy of the game: consider again the facts f and g represented in Diagram 1. f
is explanatorily prior to g, i.e. f figures in the explanation of g but not vice versa.
(Thus, there is a chain of arrows that lead from the node representing f to the one
representing g, but there is no chain of arrows leading from the g-node to the f node.) Now, suppose that in your attempt to achieve analytic consistency with the
non-obtaining of x, you are wondering whether to remove the dot representing f
or the one representing g. If you remove f , then you will lose the points attached
to f , as well as those attached to the facts that are actually explained by f . If you
remove g, you will lose the points attached to g, as well as those attached to the
facts that are actually explained by g. The latter set of facts is a proper subset of
the former. Hence, by removing f you would lose more points than by removing
g. In consequence, it seems that, ceteris paribus, removing g is the better option.
More generally, when, in reasoning hypothetically from a certain assumption, we
need to choose between holding fixed a fact g and holding fixed another fact f
that is explanatorily prior to g, we should, all other things being equal, hold f
fixed.
Given that the aforementioned three classes of truthsthe mathematical
truths, the truths about the laws, and those about the individual events in the
natural worldare ordered by explanatory priority, we can now see that the
strategic principle described in the previous paragraph is enshrined in principles
(R M ) and (R N ) of section 4.2. For what these rules tell us is that we should
choose to hold fixed the mathematical truths over holding fixed the truths about
the laws, and the truths about the laws over the truths about individual events. In
other words, we should choose to hold fixed truths from the explanatorily more
fundamental metaphysical categories of truths over holding fixed truths from the
explanatorily less fundamental categories.
properties Fred has. (For instance, the fact that Fred originated from sperm Bob
and egg cell Susie, together with certain properties of Bob and Susie (e.g. the
property of containing certain genes), might contribute to explaining why Fred
has a certain height.) But it seems to me that we cannot, in turn, explain why Fred
originated from Bob and Susie by appealing to any other properties of Fred. We
can causally explain the properties that Fred had at various times by appealing to
properties he had at earlier times, and we can explain those properties by other
properties he had at yet earlier times, and so forth. But when, in tracing back
the causal chain in this way, we have arrived at Freds property of originating
from Bob and Susie, we cannot go back any further to properties Fred had at yet
earlier times, since Fred did not exist at any earlier times (before he originated
from the two gametes). Freds property of originating from Bob and Susie is
the terminus of the causal chain that links Freds different properties, just as the
microstructure of water is the terminus of the explanatory chain that links the
different properties of water.58
Like the distinction between mathematical facts and facts about the concrete
realm, and like the distinction among the latter facts between those about what
the laws are and those about individual events, the distinction between the
essential and the accidental facts about an entity can plausibly be regarded as a
metaphysically very important distinction, a distinction that carves nature at its
joints. Our principles of weighting similarities between worlds follow the contours
of these metaphysical categories of facts about an entity. Under the influence of
(D), the weight ordering once again follows the explanatory ordering: the essential
truths about an entity are given greater weight than the accidental truths. When
deciding whether to hold fixed a certain essential truth about an entity or a
certain accidental truth, we should not choose to hold fixed the accidental truth
over holding fixed the essential truth. We should hold fixed the essential truths
if we hold fixed any truths about the entity.
(R M ) enshrines this principle, as is easy to see. (R M ) lays down that the analytically consistent worlds in which all actual mathematical and conditionalized
essential truths hold are closer than all other worlds. Now, as I said in section
2.3 (and as is proven in the appendix), this implies that, for any proposition P
that is analytically consistent with all mathematical and conditionalized essential
truths, all mathematical and conditionalized essential truths would still have been
true if P had been true. Hence, (R M ) lays down that (provided the antecedent
of a counterfactual is analytically consistent with the mathematical truths and
the conditionalized essential truths) the conditionalized essential truths are to be
held fixed. This rule in effect guarantees that the essential truths about an entity
will be held fixed as long as any true ascription of a property to the entity is held
fixed. For let P be any property of a, and let E be some essential property of the
entity a. Suppose that we are wondering what would have happened if a certain
antecedent A had been true (where A is analytically consistent with all truths
of mathematics and conditionalized essential truths). Suppose that we hold fixed
the truth that a has property P. Since (in my view) this truth entails that a exists,59
we also need to hold fixed the existence of a. But if we hold fixed the existence of
a, and also hold fixed all the conditionalized ascriptions of essential properties
to a, then we need to hold fixed all unconditionalized essential truths about a as
well. So, (R M ) guarantees that, if we hold fixed any true ascriptions of properties
to a, we need also to hold fixed all true ascriptions of essential properties to a.
As I mentioned in sections 1.2 and 3.3, it seems intuitively plausible that a
proposition that has a high degree of necessity owes this special modal status to
the fact that it is a consequence of propositions that state particularly deep and
fundamental features of the world order. The deeper these features, the higher the
degree of necessity that attaches to the proposition. The results of this section
allow us to make the ideas of depth and fundamentality a little more precise:
According to (D), what matters is the explanatory fundamentality of the facts that
make a proposition true. By and large, if the truth of a proposition flows from the
explanatorily most fundamental and basic facts of the world, then the proposition
has a high degree of necessity. I suggested above that the difference in degree of
necessity between mathematical truths, truths about the laws and truths about
individual events is grounded in the differences in explanatory depth between
these three ontological categories of propositions. The special modal force of
conditionalized essential truths derives from the relative explanatory depth of
the essential truths about the entity, compared to the accidental truths about the
same entity.
The picture I sketched in the foregoing sections is connected to a certain
strategy I endorse for explaining why our counterfactual connective is governed
by principles (R M ) and (R N ). I cannot state this explanation in this paper. I will
merely sketch the explanatory strategy and leave it to another occasion to fill in
the details.
Counterfactuals serve a number of practical purposes. They figure, e.g., in
decision-making (If I were to do such-and-such, then such-and-such would
happen). Moreover, we use them to establish claims about the explanatory
interrelations between different facts. (Suppose we know that q would not have
been the case if p had not been the case. Given certain background assumptions,
we can conclude that the fact that p plays a role in explaining the fact that
q.) In order for counterfactuals to serve their purposes adequately, the rules
that determine the closeness ordering need to have certain features, and we can
appeal to the purposes of counterfactuals to give a functional explanation of these
features. I think that it can be shown, for example, that the rules defining the
closeness ordering need to incorporate principle (C) in order for counterfactuals
to adequately serve their purpose in establishing claims about the explanatory
relations between different facts. As we saw above, this means that these rules
need to conform to the constraint stated by (D). Moreover, if counterfactual
reasoning is to be a useful reasoning strategy, then the rules that determine the
closeness ordering ought to be easy for us to handle. In order for this to be so,
they should rest on a classification of truths that is easy to apply. This will be so
if they rest on a classification that follows the contours of a conspicuous division
therefore be cotenable with P. But the propositions that are true throughout the
sphere are all and only the analytic consequences of S. The analytic consequences
of S are therefore available as ancillary premises in reasoning from any antecedent
that is analytically consistent with S.
Applications of (R M ) in hypothetical reasoning are very common. Assume
that we are reasoning from an antecedent that is analytically consistent with the
set S of all mathematical and conditionalized essential truths. Let P and Q be
two truths, and suppose that P is an analytic consequence of S, but Q is not. If P
and Q are individually but not jointly analytically consistent with the antecedent,
then at most one of them can be used as ancillary premise. We can apply (R M ) to
conclude that it is P, and not Q, that can be used. It is easy to think of ordinarylife examples in which we are using (R M ) in this way. Suppose that Susie, a shop
keeper, receives $10 from one of her customers and $20 from another. What if
the two customers together had given her, not $30, but $50? When reasoning
hypothetically from this antecedent, we cannot use both the mathematical truth
that 10 + 20 = 50, and the truth that the two customers gave Susie $10 and $20
respectively. We can regard only one of these propositions as cotenable with the
antecedent, and there is no doubt that it is the mathematical truth that we would
regard as cotenable: We would say that if Susie had received $50 from the two
customers, then at least one of them would have given her more money than he
actually did. We would not say that the sum of 10 and 20 would have been 50.
As for mathematical truths, so for true ascriptions of essential properties.
We can use a Kripkean example to illustrate the point. Imagine a situation in
which people usually use a certain odorless, colorless, tasteless liquid to quench
their thirst, brush their teeth, brew their coffee, etc. More generally, they use that
liquid in just the way in which we use water in our world. However, the liquid
they use is not composed of hydrogen and oxygen, but of some other elements.
What if a situation of this kind had obtained? We would say that, if that had been
the case, people would have used some stuff other than water to quench their
thirst, brush their teeth, and so forth. I think that we would not say that the stuff
used for these purposes would have been water, but that water would not have
been composed of hydrogen and oxygen. In this example, the antecedent (People
usually use a certain colorless, odorless, tasteless liquid that is not composed of
hydrogen and oxygen to quench their thirst, etc.) is analytically consistent with
each of the following two truths, but not with their conjunction:
Water (if it exists) is composed of hydrogen and oxygen.
Water is the colorless, odorless, tasteless liquid that people usually use to
quench their thirst, brush their teeth, etc.
At most one of these propositions can be used as ancillary premise in hypothetical
reasoning from the antecedent. As our reactions show, it is the first proposition
that we regard as permissible auxiliary premise. This is simply an application of
(R M ): it is essential to water to be a compound of hydrogen and oxygen. Therefore,
8. Conclusion
In section 1 of this paper, I argued that a good theory of modality should
combine an attractive metaphysical account of necessity with a plausible theory of
our practice of modalizing. I also described a number of ideas about necessity that
I think a good metaphysical account of modality should be able to capture: the
otherworldliness intuition, the modalist intuition, and the thought that the modal
status of a necessary proposition is grounded in its non-modal feature of flowing
from particularly deep and fundamental features of the world order. I formulated
an account of necessity according to which modal properties of propositions
come in degrees, and a propositions degree of necessity is determined by the
distance from the actual world to the closest worlds in which the proposition
is false. I argued that this account can accommodate the otherworldliness and
modalist intuitions. Drawing on the account of counterfactuals that I developed
in my (2006), I tried to show that my theory of modality also allows us to give
a plausible account of how the modal properties of propositions are grounded
in their non-modal features. Finally, I argued that my account permits us to
explain why the concept of necessity is of use to us: it facilitates counterfactual
reasoning, by singling out those propositions that play a certain special role in
that practice.
Appendix
I will prove that
(6) A deductively closed set S of true propositions has the property expressed
by the open sentence (2) if and only if the worlds that are analytically
consistent with S form a sphere around the actual world.
As I said in section 2.3 the term actual world in (6) is to be understood in the
non-rigid sense. After proving that (6) is true, I will argue that (6) is metaphysically
necessary.
First, some terminological clarifications. Let us say that two propositions P
and Q are mutually analytically consistent just in case the set containing P, Q and
all conceptual truths is narrowly logically consistent. Let us say that a world w is
analytically consistent just in case the propositions that are true in w are mutually
analytically consistent. (As I say in section 3.3, I wish to remain neutral on the
question whether every world is analytically consistent.) A world w is analytically
consistent with a set S of propositions just in case the propositions that are true
in w are jointly analytically consistent with S. A proposition Q is an analytic
consequence of P just in case Q is a narrowly logical consequence of P and the
set of all conceptual truths. I call a set deductively closed just in case it is closed
under analytic consequence. A proposition Q is cotenable with a proposition P
just in case Q is true and P ! Q.
Suppose that S is a deductively closed set of true propositions. Using the
terminology just introduced, the claim that S has the property expressed by (2)
can be reformulated by saying that S contains all and only those propositions
that are cotenable with all propositions that are analytically consistent with S.
(6) therefore amounts to the following claim: for any deductively closed set D of
true propositions,
(14) D contains all and only the propositions that are cotenable with all
propositions that are analytically consistent with D
if and only if
the worlds that are analytically consistent with D form a sphere around the
actual world.
My proof will make use of a number of assumptions that I find plausible.
Firstly, I will assume that
contingent feature of the world could their truth rest on?) Using necessitated
versions of these three assumptions, we can prove the necessitated version of (6),
i.e. the claim that
(6 N ) It is true in any metaphysically possible world w that a deductive
closed set S of true propositions has the property expressed by (2 ) if
and only if the worlds that are analytically consistent with S from a
sphere around w.
Notes
I am grateful to many philosophers for their comments on various bits and pieces
of this material, including Gordon Belot, Karen Bennett, John Burgess, Jeremy
Butterfield, Cian Dorr, Dorothy Edgington, Adam Elga, Michael Fara, Graeme
Forbes, Delia Graff, Anil Gupta, Gilbert Harman, Harold Hodes, Thomas
Hofweber, Mark Johnston, James Joyce, Thomas Kelly, Philip Kremer, Marc
Lange, Stephan Leuenberger, Eric Lormand, William Lycan, Michael McKinley,
Ram Neta, Jim Pryor, Peter Railton, Nicholas Rescher, Geoffrey Sayre-McCord,
Jonathan Schaffer, Kieran Setiya, Scott Soames, Jamie Tappenden, Fritz Warfield
and Brian Weatherson, as well as to the audiences of talks on the present material
that I gave at Princeton in December 2003, January, April and December 2004,
and at Cornell, Pittsburgh, the University of Michigan at Ann Arbor and the
University of North Carolina at Chapel Hill in January and February 2005, as
well as the students attending a graduate seminar on the material that I gave at the
University of Michigan, Ann Arbor, in the Fall Term of 2005. My greatest debt
is to Gideon Rosen, who advised me on my dissertation from which this paper
derives. I am also indebted to Neil Mehta for editorial and stylistic suggestions,
many of which I used in revising this paper, and to Jonathan Shaheen and Steve
Campbell for assistance with reading the proofs. Finally, I am grateful to Mind
for the permission to use passages from my (2006) in this paper. The relevant
passages occur in sections 2.2, 2.4, 3.3, 5.1, and 6.
1. For discussions of the point, see Kripke (1980), p. 45, Hazen (1979), pp. 320ff.,
Blackburn (1984), pp. 214f., Hale (1986), p. 77, McFetridge (1990), sct. 2, and
Rosen (1990), pp. 349ff.
2. Here is a typical example, taken from a paper by Edward Craig, in which the
author comments on the fact that conventionalist accounts of necessary truth
once enjoyed widespread popularity:
This popularity can hardly be due to any immediate plausibility or intuitive appeal. Far
from it: there is prima facie an element of paradox about Conventionalism, since [necessary]
truths ... seem to be anything but conventional, and attract attention precisely by their air
of inexorability. (1975, p.1, my emphasis)
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
38.
39.
40.
41.
42.
where Hesperus, Phosphorus and Alpha Centauri are understood as directly referential expressions?
I am thinking of propositions as structured entities. The proposition expressed
by a sentence containing a directly referential expression for the entity e is an
object-dependent proposition that contains e as one of its constituents. The objectdependent proposition expressed by (1) contains Hesperus (i.e., Phosphorus) as
constituent twice over. It is therefore of the logical form a = a. It seems plausible
to say that the proposition is true in virtue of having this logical form, and is
in this sense a logical truth. As a logical truth, it is a logical consequence, and
therefore an analytic consequence, of every set of propositions.
Consider next the proposition expressed by (2). This proposition is of the logical
form (a = b), but not of the form (a = a) (since there is no constituent that
occurs twice in it). Every proposition that instantiates the first form but not the
second is a true statement of non-identity, and it seems plausible to me to say
that the proposition under consideration is true in virtue of the relevant two facts
about its logical form. The proposition is a logical truth in the sense of being true
in virtue of its logical form. As a logical truth, it is a logical consequence, and
therefore an analytic consequence, of every set of propositions.
43. This section consists mainly of material from my (2006).
44. For reason which I cannot expand on here (but which I discuss in Kment (MS)),
I think that this account of the closeness relation is a little bit too simple. But I
think that it is a good approximation to the truth, and it can serve as a perfectly
good working account for my purposes.
59.
60.
61.
62.
63.
64.
65.
66.
67.
(viz. that of originating from Fred and Susie) is itself an essential property of
Fred.) My thesis is merely that no essential property of an entity can be explained
by an accidental property of the same entity, whereas accidental properties can
often be explained by essential properties. There is a clear explanatory asymmetry
between the essential and the accidental facts about an entity.
As I mentioned in footnote 39, I believe that no entity has properties in any
metaphysically possible world where it does not exist.
Principle (D) constrains what the principles of weighting can be like. But it does
not entail (R M ) or (R N ). There are other sets of principles that we could have used
to weight similarities, and which also conform to the constraint imposed by (D).
Nonetheless, I think that there is something special about a system of rules
that includes (R M ) and (R N ). Every possible system of weights divides up the
facts about our world into different classes and orders these by how weightily it
contributes to the closeness of another world if it shares with the actual world facts
in the relevant classes. Some possible sets of rules rely on classifications of facts
that carve nature at its joints, others rely on more gerrymandered classifications.
As I said, (R M ) and (R N ) are rules of the former kind. They rely on metaphysically
very significant and central classifications of facts, perhaps the most significant
metaphysical distinctions between facts which there are. I conjecture that, of all
the possible principles of weighting that conform to the constraints imposed by
(D), there are none that rest on a metaphysically deeper and more important
classification of facts. In consequence, I surmise that (R M ) and (R N ) optimally
satisfy the desideratum of being easy to apply while conforming to the constraint
imposed by (D).
(1947).
Scts. 57.
This specification of the closeness relation is cast in the same form as Lewiss
well-known formulation in his (1986b), pp. 47f.
Jackson (1977), p. 9.
If the antecedent is analytically inconsistent and we allow for analytically inconsistent worlds, then the closest antecedent-worlds are analytically inconsistent.
In that case, there may be some restriction on the principles of inference that
can be used in the hypothetical reasoning, since some of the inference rules that
are actually valid might be invalid in the closest antecedent-worlds. But if the
antecedent is analytically consistent, then all actually valid principles of inference
are applicable without restriction.
From the perspective of the closeness account, the process of hypothetical
reasoning can be described as follows: Suppose that we want to find out whether
P ! Q is true. This requires us to determine whether the P-worlds closest to
the actual world are Q-worlds. We can do so by applying both our background
knowledge about the actual world and the rules defining the closeness relation to
show that certain true propositions are also true in the closest P-worlds, and by
then reasoning from P and the relevant truths in order to determine whether they
entail Q. If they do, then we can conclude that Q is true in the closest P-worlds
and that P ! Q is therefore true.
This definition is slightly different from the one Goodman originally proposed
((1955), p. 15). On his definition, a proposition P is cotenable with an antecedent
References
Adams, E. (1975), The Logic of Conditionals, Dordrecht (Reidel)
Adams, R. M. (1981), Actualism and Thisness, Synthese 57, pp. 342
Bennett, J. (1984), Counterfactuals and temporal direction, Philosophical Review 93, pp. 5791
Bennett, J. (2001), On Forward and Backward Counterfactual Conditionals, in: Reality and
Humean Supervenience, ed. G. Preyer and F. Siebelt, Lanham (Rowman and Littlefield),
pp. 177202
Bennett, J. (2003), A Philosophical Guide to Conditionals, Oxford (Clarendon)
Blackburn, S. (1984), Spreading the Word. Groundings in the Philosophy of Language, Oxford
(Clarendon)
Craig, E. (1975), The Problem of Necessary Truth, in: Meaning, Reference and Necessity, ed.
S. Blackburn, Cambridge (Cambridge UP), 131.
Edgington, D. (2003), Counterfactuals and the Benefit of Hindsight, in: Causation and
Counterfactuals, ed. P. Dowe and P. Noordh of, London (Routledge)
Fine, K. (1977), postscript to A. N. Priors Worlds, Times, and Selves, London (Duckworth),
1168
Fine, K. (1994), Essence and Modality, Philosophical Perspectives 8, pp. 116
Fine, K. (2002), The Varieties of Necessity, in: Imagination, Conceivability and Possibility, ed.
J. Hawthorne and T. Gendler, Oxford (Clarendon), pp. 25381
Forbes, G. (1985), The Metaphysics of Modality, Oxford (Clarendon)
Goodman, N. (1947), The problem of counterfactual conditionals, Journal of Philosophy 44,
pp. 11328
Goodman, N. (1955), Fact, Fiction and Forecast, Indianapolis (Bobbs-Merrill).
Hale, B. (1986), The Compleat Projectivist, Philosophical Quarterly 36, pp. 6584
Hazen, A. (1979), Counterpart-Theoretic Semantics for Modal Logic, Journal of Philosophy
76, pp. 31938
Hiddleston, E. (2005), A Causal Theory of Counterfactuals, Nous 39(4), pp. 632657
Hill, C. (forthcoming), Modality, Modal Epistemology, and the Metaphysics of Consciousness,
in S. Nichols (ed.), The Architecture of the Imagination: New Essays on Pretense,
Possibility, and Fiction, Oxford (UP)
Jackson, F. (1977), A Causal Theory of Counterfactuals, Australasian Journal of Philosophy
55, pp. 321
Kment, B. (2006), Counterfactuals and Explanation, Mind 115, pp. 261310
Kment, B., MS, The Similarity Account of Counterfactuals
Kment, B. (dissertation), A Theory of Counterfactuals and a Counterfactual Theory of Necessity,
Princeton 2006
Kripke, S. (1980), Naming and Necessity, Cambridge, Mass. (Harvard UP)
Lange, M. (1999), Laws, Counterfactuals, Stability, and Degrees of Lawhood,Philosophy of
Science 66, 24367
Lange, M. (2000): Natural Laws in Scientific Practice. New York: Oxford University Press.
Lange, M. (2004), A Note on Scientific Essentialism, Laws of Nature, and Counterfactual
Conditionals, Australasian Journal of Philosophy 82, 227241
Lange, M. (2005), A Counterfactual Analysis of the Concepts of Logical Truth and Necessity,
Philosophical Studies 125, pp. 277303
Lewis, D. K. (1973), Counterfactuals, Oxford (Blackwell)